1、pip install tensorflow-gpu==2.1.0 --user（pip install tensorflow-gpu==2.1.0 -i https://pypi.tuna.tsinghua.edu.cn/simple --user）
2、安裝VC_redist.x64.exe：
安裝
2、import tensorflow出錯：
Could not load dynamic library 'cudart64_101.dll'; dlerror: cudart64_101.dll not found
解決辦法：https://cn.dll-files.com/cudart64_101.dll.html
下載cudart64_101.zip
解壓后把dll文件拷貝到：C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\bin
3、import成功：
>>> import tensorflow
2020-03-23 15:37:43.258666: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library cudart64_101.dll
>>> tensorflow.__version__
'2.1.0'
4、升級cuda到10.1版本：https://developer.nvidia.com/cuda-downloads 下載：cuda_10.1.105_418.96_win10，自定義安裝。
安裝成功后，驗證：
import tensorflow as tf
#驗證cuda
a = tf.test.is_built_with_cuda()
print(a)
physical_devices = tf.config.experimental.list_physical_devices('GPU')
print(physical_devices)
5、升級cudnn到7.6.5：
安裝cudnn7.6.5 for cuda10.1 for windows，下載：cudnn-10.1-windows10-x64-v7.6.5.32.zip，https://developer.nvidia.com/rdp/cudnn-download
把bin、include、lib里面的文件分別復制到cuda安裝目錄對應子目錄。